Table 3.

Final model of multiple linear regression analysis results on predictors of the normalized changes of IED_errors after TCC training.

	Unstandardized coefficients		Standardized coefficients		Adjusted R2	Adjusted R2 change	F	p-Value
	B	Standard error	Beta	p-Value
Constant	13.560	3.663		0.002*
Covariates
Age	−0.066	0.025	−0.547	0.018*
Sex	−0.831	0.208	−0.713	0.001*
Education	−0.107	0.040	−0.491	0.018*
GFApre of the PSTP loop	−13.871	4.610	−0.620	0.009*	0.507	0.265	9.054	0.009*

GFA_pre, generalized fractional anisotropy at baseline; IED_errors, the number of total errors of the Intra/Extra-dimensional set shift test; PSTP, prefronto-striato-thalamo-prefrontal; TCC, Tai Chi Chuan group. Normalized change of IED_errors = (post-test IED_errors – pre-test IED_errors)/pre-test IED_errors.

The stepwise method was applied, using the criteria of p(F) < 0.05 as the probability-to-enter and p(F) ≧ 0.10 as probability-to-remove. Age, sex, and education were entered as covariates. The GFA_pre values of the PSTP loop and prefronto-parietal/occipital fiber group were originally entered into the regression analysis, but the GFA_pre of the prefronto-parietal/occipital fiber group was removed in the final model based on the set criteria.

^*p < 0.05.